ruby-kafka

ruby-kafka copied to clipboard

This PR prevents the example "create partitions" from failing like https://app.circleci.com/pipelines/github/zendesk/ruby-kafka/221/workflows/7d51470f-c102-4b30-b2e0-68c618eeb1d2/jobs/6799?invite=true#step-104-122.

As there is a time lag for Kafka's metadata to be updated, Kafka::Client#partitions_for is needed to be executed until the metadata is updated to check the latest partition count. However, Kafka::Cluster never fetches new metadata until it considers the current metadata stale. So, this PR adds the method Kafka#refresh_metadata in order to force Kafka::Cluster to fetch metadata. The option helps in production code as well to ensure we get the latest metadata.